Benchmarks · /benchmarks/bridgebench-dgx-throughput

DGX throughput

Name: DGX throughput
Creator: BridgeBench

BridgeBench DGX Spark local-model throughput.

Source · BridgeBench
Version · bridgebench snapshot 2026-05-15
Scores · 4

Test details

Verified but agingThis is an efficiency signal, so it belongs beside quality rather than being mistaken for quality.

source

BridgeBench

metric

tok/s (tok/s)

judge

Speed / cost

direction

higher better

group id

bridgebench_dgx_throughput_2026_05

domain

Coding

What it measures vs what it misses

✓ Measures

Tokens per second on the DGX Spark local hardware setup.

✗ Misses

Hosted serving throughput. Output correctness.

Why this countsIt tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.Same-test ruleThis percentile only compares models inside the exact benchmark/version group shown here. It is not a universal score.What it missesIt does not fully capture repo-scale iteration, IDE ergonomics, or long debugging loops.

Leaderboard · this benchmark version

#1 · GPT-OSS 120B

BB · Undated

verified runtimeexact directBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 100%
Last updated: archived
Eligibility: historical_model
Identity: exact (1.00)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

41.9 tok/s

#2 · gemma-4-31b

BB · Undated

Source label: gemma-4-31b-local

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 66.7%
Last updated: archived
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

16.5 tok/s

#3 · Qwen 3.5 27B local

BB · Undated

verified runtimeexact directBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 33.3%
Last updated: archived
Eligibility: Local DGX Spark configuration tracked separately from hosted general rankings.
Identity: exact (1.00)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

11.1 tok/s

#4 · Mistral Small 4

BB · Undated

Source label: mistral-small-4-local

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 0%
Last updated: archived
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

4.7 tok/s

Benchmarks · /benchmarks/bridgebench-dgx-throughput

DGX throughput

BridgeBench DGX Spark local-model throughput.

Source · BridgeBench
Version · bridgebench snapshot 2026-05-15
Scores · 4

Test details

Verified but agingThis is an efficiency signal, so it belongs beside quality rather than being mistaken for quality.

source

BridgeBench

metric

tok/s (tok/s)

judge

Speed / cost

direction

higher better

group id

bridgebench_dgx_throughput_2026_05

domain

Coding

What it measures vs what it misses

✓ Measures

Tokens per second on the DGX Spark local hardware setup.

✗ Misses

Hosted serving throughput. Output correctness.

Leaderboard · this benchmark version

#1 · GPT-OSS 120B

BB · Undated

verified runtimeexact directBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 100%
Last updated: archived
Eligibility: historical_model
Identity: exact (1.00)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

41.9 tok/s

#2 · gemma-4-31b

BB · Undated

Source label: gemma-4-31b-local

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 66.7%
Last updated: archived
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

16.5 tok/s

#3 · Qwen 3.5 27B local

BB · Undated

verified runtimeexact directBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 33.3%
Last updated: archived
Eligibility: Local DGX Spark configuration tracked separately from hosted general rankings.
Identity: exact (1.00)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

11.1 tok/s

#4 · Mistral Small 4

BB · Undated

Source label: mistral-small-4-local

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/dgx-spark
Percentile: 0%
Last updated: archived
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-dgx-throughput.

4.7 tok/s

DGX throughput

Test details

What it measures vs what it misses

✓ Measures

✗ Misses

Leaderboard · this benchmark version

Loading benchmark evidence.

DGX throughput

Test details

What it measures vs what it misses

✓ Measures

✗ Misses

Leaderboard · this benchmark version